- From: Al Gilman <Alfred.S.Gilman@IEEE.org>
- Date: Thu, 22 Feb 2007 12:56:37 -0500
- To: semantic-web@w3.org
- Cc: Lisa Seeman <lisa@ubaccess.com>, wai-liaison@w3.org
Dear friends of semantics, ** A little background In WAI-ARIA we *are* as of now using @role for things that have reference explanations that are accessible by QName. And the 'reference explanations' in our present approach are classes in an OWL ontology. We are not sure about how constraints, e.g. cardinality, in the ontology affects conformance once an instance names that class in its @role value. ** Two theories Here are two ways of understanding how it works: Theory A: Owl cardinality constraint can only refer to usages inside RDF. I don't think we can legally use it to constrain the usage of native mark up which it abstracts. Theory B: The syntax given in our XHTML 1.1 dialect defines patterns that match, that specialize the classes in the ontology. (There is, in principle, a binding that says what patterns in the syntax are specializations of what patterns in the abstract model.) Instances of those patterns in a WAI-ARIA-conforming document in that syntax satisfy the constraints given in the abstract OWL pattern. The use of one of our role definitions in OWL as the value of @role in an XHTML 1.1-extended document imports the assertions, including constraint assertions, that are written in the OWL and says they are applicable to this element instance in the XHTML. Rationale -- why do we care? People make interactive widgets by scripting all sorts of syntactic elements. WAI-ARIA is a way to live with that. But an interactive element where the user can do something requires (for accessiblity) a label giving a notion of what it does. In legacy HTML without the scripted behavior, we have guidelines linking this to the INPUT element. In XForms, it's enforced by the specification. But in the wild west of AJAX, shall we say, we seek to hang this rule on a Role. We can't make the population constraint that there has to be an @aaa:labeledby attribute in the syntax because it cuts across element types. But we can say in the role template that something of this nature has to be there, because it is specific to certain roles. The question is, is it binding? Inheritance here, if it works, is transitive: The grand-specific inherits constraints from the grand-generic and the grand-generic inherits instances from the grand-specific. here: - grand-specific is a class in our ontology - mid-specific = mid-generic is some pattern in the DOM of the syntax - grand-specific is a document instance that aspires to conform to our spec. We can't just replicate the population constraints in the DTD because they are cross-attribute of the form (by example): WHEN attribute @role has value exampleQName THEN attribute @labeledby MUST occur at least once. ** recap The question is: Can we, or can't we, constrain instances of our DOCTYPE from constraints in the OWL that aren't replicated or narrowed in the DTD. Al
Received on Thursday, 22 February 2007 18:33:00 UTC